Virtualization is everywhere. It's not just because physical servers are converted to virtual machines, but that the physical host machines can be virtualized as well, using advanced solutions from companies like Cisco or Hewlett-Packard.
Currently, Cisco's Unified Computing System is the market leader in converged infrastructure systems. HP Blade System Matrix and IBM's PureSystems offer similar solutions. In this article you'll read about unified computing systems (UCS) in general, so when UCS is mentioned, it is not just about the Cisco solution, but about the solution in general, which includes solutions of other vendors with different product names.
The basic concept behind a unified computing platform is to integrate as much as possible into one single box. That means that the computer hardware, the storage area network (SAN) connections and the networking infrastructure are all in the same solution with the goal to make management easier. Typically, blade systems or rack-mounted servers can be used as the compute nodes. The blades or servers are not addressed as separate entities, but as building blocks that can be used to compose a computing entity. By taking that approach, the hardware itself is virtualized and makes additional redundancy possible. Virtualizing the hardware also makes it more difficult to find which pieces of hardware are currently hosting specific services.
A second important part of the solution is the integrated network. IT shops tend to use a 10 Gbps integrated network that connects to external servers and users and to the SAN. Even if in theory the SAN can be integrated in the UCS, in practice it is often external to the unified computing solution. In Cisco's UCS, Fibre Channel over Ethernet is used to make it possible to use one integrated switch to address the user network as well as the storage network.
An integrated management interface allows the administrator to code the individual hardware components into the functional solution and to build a solution that shuts down failing components. This offers fencing, which is often required in high-availability clusters to make sure that failing nodes are shut down properly.
Converged infrastructure makes it easy to deploy virtualized solutions in an efficient manner.
Sander van Vugt
Integrating virtualization in a UCS
Converged infrastructure makes it easy to deploy virtualized solutions in an efficient manner. Because everything is in one box, the amount of rack space that is required is limited, which makes the solution relatively cost efficient. In addition, if the solution has integrated support for virtualization, the virtual environment can be managed from the integrated management interface, which is the case for the Cisco UCS. That makes it possible to optimize the efficiency of the solution and to have the platform do something in case a hardware component goes down.
The management solution doesn't have to be limited to the UCS. Cisco Integrated Management Controller, for instance, also makes it possible to manage external servers from the same interface.
Alternatives and drawbacks
Solutions like Cisco's UCS or HP's Blade System Matrix are becoming popular in data centers for optimizing management and the use of available resources. There are a few drawbacks though. The most important drawback is that it is difficult to integrate hardware from other vendors. In Cisco's UCS with an integrated Nexus switch, it all works perfect if everything in the IT shop is on Cisco Nexus. Of course, non-Cisco switches can be integrated in the switching topology, but they cannot be managed from the same interface and they cannot rely on the integrated fault tolerance of a UCS. Implementing an integrated solution automatically means increasing the shop's dependency on one vendor and makes it difficult, if not impossible, to use third-party hardware.
That is just the price to be paid for increased efficiency in the data center. It is inherently inefficient to use hardware from different vendors and maximize management efforts. That also goes for other solutions where data center efficiency is optimized, even if these solutions take a completely different approach to efficiency such as the purchase of a midrange or mainframe computer.
ABOUT THE AUTHOR: Sander van Vugt is an independent trainer and consultant based in the Netherlands. He is an expert in Linux high availability, virtualization and performance, and has completed several projects that implement all three. He is also the writer of various Linux-related books, such as Beginning the Linux Command Line, Beginning Ubuntu Server Administration and Pro Ubuntu Server Administration.